Comments on: Closures Part I http://www.mirandabanda.org/cogblog/2008/06/07/closures-part-i/ Speeding Up Croquet and Squeak with a new open-source VM from Teleplace Tue, 27 Jul 2010 17:32:30 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: Mathueu Suen http://www.mirandabanda.org/cogblog/2008/06/07/closures-part-i/comment-page-1/#comment-27 Mathueu Suen Mon, 16 Jun 2008 13:51:48 +0000 http://cogblog.mirandabanda.org/?p=4#comment-27 Hi Eliot I am wondering why you do not use the Anthony Hannan’s compiler. It seem to me that all the feature you want is inside. <em>I'm a slow learner and find picking up new things hard. I also know exactly what I want to do with Cog. So it is easier for me to use the existing compiler that I know very well. It took me two weeks to modify the compiler to do closures. It probably would have taken me that amount of time to get to grips with Anthony's compiler.</em> As far as I understand is because of optimization that need to be done at the compiler level. I would be very interest if you explain in detail why. <em>I hope I will in future posts. The issue is to do with how the Vm fools the image into thinking there are always contexts and only contexts when in fact the Vm uses a much more conventional and much faster stack organization.</em> One thing that can be done with the compiler of Anthony is rewriting the BytecodeGenerator so you can change the bytecode generated to create/access ClosureEnvironment. <em>Well I think I did the same thing, just on what for me is more familiar ground. No disrespect to Anthony's compiler. And I had fun doing it myself :)</em> The post is very interesting and I am interest to see the post to come about context-to-stack mapping and the scope analysis :) . Thanks <em>Thank you! I'll try not to keep you waiting! Eliot</em> Hi Eliot

I am wondering why you do not use the Anthony Hannan’s compiler. It seem to me that all the feature you want is inside.

I’m a slow learner and find picking up new things hard. I also know exactly what I want to do with Cog. So it is easier for me to use the existing compiler that I know very well. It took me two weeks to modify the compiler to do closures. It probably would have taken me that amount of time to get to grips with Anthony’s compiler.

As far as I understand is because of optimization that need to be done at the compiler level. I would be very interest if you explain in detail why.

I hope I will in future posts. The issue is to do with how the Vm fools the image into thinking there are always contexts and only contexts when in fact the Vm uses a much more conventional and much faster stack organization.

One thing that can be done with the compiler of Anthony is rewriting the BytecodeGenerator so you can change the bytecode generated to create/access ClosureEnvironment.

Well I think I did the same thing, just on what for me is more familiar ground. No disrespect to Anthony’s compiler. And I had fun doing it myself :)

The post is very interesting and I am interest to see the post to come about context-to-stack mapping and the scope analysis :) .
Thanks

Thank you! I’ll try not to keep you waiting!

Eliot

]]> By: Jecel Assumpção Jr http://www.mirandabanda.org/cogblog/2008/06/07/closures-part-i/comment-page-1/#comment-24 Jecel Assumpção Jr Fri, 13 Jun 2008 22:18:33 +0000 http://cogblog.mirandabanda.org/?p=4#comment-24 In the example that follows "This way round works as intended:" I think you meant to put the "n" before the rest of the expression as indicated in the paragraph after that. Right now the two versions are identical. It took me quite a while to convince myself of that :-) <em>Thanks Jecel! Good catch. I completely reformatted the code, going to generated HTML: instead of using < pre > and I screwed up second time around. So now this way round it works :-) The fact that I can edit my previous attempts is convenient. Corrupting even. Eliot</em> In the example that follows “This way round works as intended:” I think you meant to put the “n” before the rest of the expression as indicated in the paragraph after that. Right now the two versions are identical. It took me quite a while to convince myself of that :-)

Thanks Jecel! Good catch. I completely reformatted the code, going to generated HTML: instead of using < pre > and I screwed up second time around. So now this way round it works :-)

The fact that I can edit my previous attempts is convenient. Corrupting even.

Eliot

]]> By: Anthony Lander http://www.mirandabanda.org/cogblog/2008/06/07/closures-part-i/comment-page-1/#comment-22 Anthony Lander Thu, 12 Jun 2008 16:00:10 +0000 http://cogblog.mirandabanda.org/?p=4#comment-22 Eliot, Teaching as you are building is a wonderful idea, and we (the interested readers) are the lucky beneficiaries. Thank you for sharing with us. -Anthony <em>I feel so privileged to have this fantastic context within which to work on Cog. Great blogging software, an affordable hosting service, an open source project to work on, a great product to contribute to and great colleagues. I feel like the Ferrari test driver I saw in a program on car design who, on getting out of a 456 he was testing to destruction doing doughnuts until a radiator hose broke, said "I am the Ferrari test driver. I have the best job in the world.". Later in the program you can guess what the Lamborghini test driver said :-) Eliot</em> Eliot,

Teaching as you are building is a wonderful idea, and we (the interested readers) are the lucky beneficiaries. Thank you for sharing with us.

-Anthony

I feel so privileged to have this fantastic context within which to work on Cog. Great blogging software, an affordable hosting service, an open source project to work on, a great product to contribute to and great colleagues. I feel like the Ferrari test driver I saw in a program on car design who, on getting out of a 456 he was testing to destruction doing doughnuts until a radiator hose broke, said “I am the Ferrari test driver. I have the best job in the world.”. Later in the program you can guess what the Lamborghini test driver said :-)

Eliot

]]>